Accounting for Dynamic Fluctuations across Time when Examining fMRI Test-Retest Reliability: Analysis of a Reward Paradigm in the EMBARC Study.

Henry W Chase,Richelle Stiffler,Jay C Fournier,Crystal Cooper,Madhukar H Trivedi,Maurizio Fava,Sarah Weyandt,Tom Carmody,Haris Aslam,Melvin G Mcinnis,Mary L Phillips,Thilo Deckersbach,Tsafrir Greenberg,Patrick J Mcgrath,Marisa Toups,Carlos R Zevallos,Scott Peltier,Phillip Adams,Jorge Almeida ,Ramin V Parsey ,Myrna M Weissman ,María A Oquendo ,Benji T Kurian

doi:10.1371/journal.pone.0126326

Abstract

Longitudinal investigation of the neural correlates of reward processing in depression may represent an important step in defining effective biomarkers for antidepressant treatment outcome prediction, but the reliability of reward-related activation is not well understood. Thirty-seven healthy control participants were scanned using fMRI while performing a reward-related guessing task on two occasions, approximately one week apart. Two main contrasts were examined: right ventral striatum (VS) activation fMRI BOLD signal related to signed prediction errors (PE) and reward expectancy (RE). We also examined bilateral visual cortex activation coupled to outcome anticipation. Significant VS PE-related activity was observed at the first testing session, but at the second testing session, VS PE-related activation was significantly reduced. Conversely, significant VS RE-related activity was observed at time 2 but not time 1. Increases in VS RE-related activity from time 1 to time 2 were significantly associated with decreases in VS PE-related activity from time 1 to time 2 across participants. Intraclass correlations (ICCs) in VS were very low. By contrast, visual cortex activation had much larger ICCs, particularly in individuals with high quality data. Dynamic changes in brain activation are widely predicted, and failure to account for these changes could lead to inaccurate evaluations of the reliability of functional MRI signals. Conventional measures of reliability cannot distinguish between changes specified by algorithmic models of neural function and noisy signal. Here, we provide evidence for the former possibility: reward-related VS activations follow the pattern predicted by temporal difference models of reward learning but have low ICCs.

Highlights

It is essential to improve the test-retest reliability of the blood oxygenation level dependent (BOLD) signal both to provide a deeper understanding of individual differences in context-dependent neural responses, as well as a meaningful interpretation of functional neural circuitry changes across time
As there were no further factors of theoretical interest, we focused solely on the first factor, and used this as a way to divide the healthy control (HC) cohort into ‘high signal to noise ratio (SNR)’ (n = 18) and ‘low SNR’ subgroups (n = 19): all participants with negative loadings were placed in the high SNR group and all with positive loadings were placed in the low SNR group
Significant right ventral striatum (VS) prediction error (PE)-related activation was observed at time 1 (t = 5.65, p

Summary

Introduction

It is essential to improve the test-retest reliability of the blood oxygenation level dependent (BOLD) signal both to provide a deeper understanding of individual differences in context-dependent neural responses, as well as a meaningful interpretation of functional neural circuitry changes across time. As might be expected, the literature as a whole seems to align somewhere between the two, with a modest (‘fair’) reliability: the pooling of studies reporting Intra-Class Correlations (ICCs), a conventional metric of data reliability, yielded an average ICC of around 0.5 [3] Such levels of reliability would be unacceptable in many other fields of scientific investigation, and this has tempered optimism about the use of functional magnetic resonance imaging (fMRI) to provide meaningful insight into individual differences [4]. Emotion-related amygdala activation was generally unreliable, as assessed by a conventional measure of reliability (ICCs: (see [2, 10, 11]), and was not consistently improved by alternative modeling strategies This was observed despite the capability of the procedure to reveal clinically relevant individual differences in amygdala activation [12]. Among several possible explanations for this discrepancy, one likely explanation is that as predicted by previous empirical [13, 14] and theoretical conceptions of amygdala function [15, 16], amygdala activation may change dynamically during the experimental paradigm itself

Methods

Results

Discussion

Conclusion